CRM Migration Plan: A Practical Guide for Sales Leaders
A CRM migration plan is the document that decides who owns what, which data moves, how you prove it arrived intact, and what happens if cutover goes wrong. It is not an import schedule. If your plan fits on a CSV tab, you don’t have a plan yet.
We’ve seen migrations that were technically “complete” while the forecast was off by a quarter, because the records moved but the relationships between them didn’t. Real talk: the import is the easy part. The hard part is agreeing on definitions, owners and acceptance tests before anyone touches production.
Below is the plan we’d want on the table if you were a Head of Sales or CEO about to switch systems, with a phase checklist you can hand to your RevOps lead.
A solid plan is a controlled business and technology change. It covers goals, scope, owners, timeline, data audit and cleansing, field and object mapping, destination configuration, testing, cutover, validation, training and post-launch optimization. HubSpot’s 2026 process overview frames it with goals, constraints such as the freeze window and rollback trigger, and success criteria such as record counts, accuracy rates and user validation tests.
Salesforce’s own migration guidance defines the job as moving data, configurations and business processes, not only records. That distinction matters. Your workflows, reports, permissions and integrations are part of what you’re migrating, and most of them are where the surprises live.
The sections that follow turn that into phases.
Before a vendor demo, write down the three business problems the new CRM must solve. “Our reps hate it” is a symptom. “Forecast accuracy is poor because stage definitions differ by team” is a problem a migration can fix, and one it can also reproduce if you move the mess intact.
Name the people first:
This isn’t bureaucracy. Unresolved ownership turns a mapping exception or a duplicate dispute into a political delay instead of a technical decision. Settle decision rights now and you’ll save weeks later.
Before you pick a cutover date, list what exists. The Salesforce migration guide points to contacts, accounts, opportunities, activities, products, quotes, custom fields, workflows, reports, permissions and integrations. We’d add the ones people forget: forms, calendars, billing links, telephony, marketing automation and ERP connections.
Then classify each item into one of four buckets:
| Bucket | Meaning | Typical examples |
|---|---|---|
| Migrate | Moves as is, or with light cleanup | Active accounts, open opportunities, current contacts |
| Archive | Kept outside the CRM, retrievable | Closed-lost deals older than your reporting window, old email logs |
| Rebuild | Logic moves, implementation doesn’t | Lead routing, approval workflows, scoring |
| Retire | Dies with the old system | Unused custom fields, dead reports, abandoned automations |
The highest-leverage decision in the whole project is what not to move. Every field you retire is a field nobody has to map, test or train on. Real talk: if nobody can tell you who uses a report, it goes in Retire and you wait to see who complains.
This is also the moment to settle how your sales process should look in the new system. If your stages and exit criteria live in people’s heads, our six-stage process documentation template will get them on paper in about 30 days, which makes configuring the destination far less arbitrary.
Write acceptance criteria now, while nobody is defending a deadline. Good ones are measurable:
HubSpot’s practical process overview makes the same point: a migration can look successful while sales reporting is wrong, so reconcile pipeline, associations and field-level samples instead of trusting row totals. A hundred percent of rows can arrive with the wrong parent account on a fifth of them.
If you don’t yet have a consistent set of revenue metrics to reconcile against, fix that in parallel. Our write-up on RevOps metrics that predict revenue lays out formula, source, cadence and owner for each, which is exactly what a validation checklist needs.
Cleansing is where the project is won. Plan for:
HubSpot’s import checklist sequences it in the same order: export or connect, cleanse, map objects and fields, prepare destination fields, build association-ready files, run a test import, and review. Notice that the test import comes before the full load.
One more thing leaders skip: account matching. If leads can’t be tied to the right account at import, you’ll misroute pipeline from day one. See how we approach lead-to-account matching before you finalize your matching keys.
Build one mapping document, in a spreadsheet, with a row per source field: source object and field, destination object and field, data type, transformation rule, required or not, owner, and migrate/archive/rebuild/retire status. Nothing gets loaded that isn’t in the sheet.
Mapping isn’t just fields. Decide the system of record for each object. If your CRM talks to an ERP, the answer differs by object, and “sync everything both ways” is the sentence that should make you nervous. Microsoft’s initial synchronization guidance is a good example: it explains that single-threaded and multithreaded tables use different load directions, so sequence and direction are not interchangeable. Microsoft also states that leads, opportunities and quotes stay in Dataverse and don’t synchronize to Finance and Operations when there’s no downstream accounting tied to them. Don’t assume every sales object belongs in the ERP migration.
The practical ask for you as the executive: demand an object-by-object system-of-record matrix from whoever runs the project.
If marketing automation feeds your CRM, treat that mapping as its own workstream. Bad field maps break marketing automation integrations quietly, and you’ll find out when lead volume looks fine but nothing gets routed.
Set up the destination before you load a record: users, roles, permission sets, required fields, picklists, record types, pipelines and stages, and the automations you flagged as Rebuild. Then load in dependency order:
Run a representative pilot first: a slice that includes your messiest accounts, not your cleanest. Review mappings, associations and engagement data against the source, fix the rules, and repeat until the pilot passes your acceptance criteria. Two clean pilots beat one lucky one.
Cutover is a rehearsed event, not an overnight improvisation. A safe sequence looks like this:
Rollback triggers should be numbers, not feelings: for example, pipeline variance above your tolerance, a failed integration test on lead intake, or a permissions breach. Write them down before the stress arrives.
On backups, CISA’s Cyber Essentials guidance recommends automated backups, protected off-site or cloud copies, encryption and at least seven days of incremental rollback capability, and it stresses testing recovery. Restore your source backup into a sandbox at least once. If you can’t, you don’t have a rollback.
Big bang or phased? A single-team, low-integration setup can often cut over in one weekend. Multiple regions, a heavy ERP link or many custom automations favor phasing by team or region. Choose based on how many integrations you’d have to debug at once.
If your database includes contacts in the EU, and for many B2B tech companies it does, the export file is personal-data processing. GDPR Article 5 sets principles such as data minimization, accuracy and storage limitation, and Article 32 covers security of processing. Obligations depend on your role and context, so take the specific migration to your privacy counsel.
The practical effect on the plan:
If outbound is part of your motion, also read our piece on audit-ready records for B2B cold email under GDPR, since consent and source records are exactly the fields people forget to map.
A migration is not done at go-live. Plan two to four weeks of hypercare: a daily triage of defects, a visible issue log, office hours, and a short list of metrics you watch every morning (lead intake, opportunity creation, stage movement, forecast variance).
Train by role, using real accounts, not demo data. Reps need to know where their pipeline went; managers need their forecast view; finance needs the closed-won feed. Keep the old system read-only for a defined period, then archive it deliberately instead of letting it drift.
Then revisit the downstream handoffs. Your sales-to-customer-success handoff probably lives partly in CRM fields and tasks, and it’s often the first thing to break quietly.
We come at this as practitioners who’ve run revenue teams in tech, not as people who learned CRM from a vendor deck. Here’s what we believe.
First, the CRM is the visible part of a revenue system. Underneath sit your ICP, roles, stage definitions, routing rules, governance and reporting. If those are fuzzy, a new CRM gives you the same fuzziness with a cleaner interface. That’s why we prefer end-to-end system design over isolated fixes, and why a migration is the best moment you’ll get to redesign.
Second, nobody should be a hero on cutover weekend. If success depends on one admin staying up all night, the plan is thin. Structure beats heroics: rehearsals, written triggers, named owners.
Third, AI is no reason to wait. It may change data entry, scoring, forecasting and search, but it raises the value of clean, governed, permissioned data. Design the target system with clear data ownership, automation you can explain, and a human approving consequential decisions.
Finally, be skeptical of any promise to “just move everything.” The right answer is almost always to move less, and move it better.
— Antony
If you’re weighing a move and aren’t sure your process, data model and team structure are ready for one, a Sales Workflow Audit or Revenue System Diagnostics is a sensible first step. We look at how your revenue system actually runs, then help you design what the new CRM should hold before anyone configures it. For a wider view of how the CRM fits with the rest of your tools, start with our guide to the B2B sales tech stack.
| Phase | Must be true before moving on |
|---|---|
| 0. Decide | Business problems written, sponsor and migration lead named |
| 1. Inventory | Every object, field, workflow, report and integration classified |
| 2. Success criteria | Acceptance thresholds and rollback triggers documented |
| 3. Cleanse | Duplicates resolved, picklists standardized, survivorship rules signed |
| 4. Map | Mapping sheet complete, system of record set per object |
| 5. Pilot | Representative pilot passes reconciliation |
| 6. Cutover | Rehearsal done, backup restored in a test, freeze announced |
| 7. Hypercare | Daily metrics watched, users trained by role, old system archived on plan |
It depends on data volume, number of integrations and how much you rebuild, so we won’t give you a universal number. A single team with clean data and few integrations is a very different project from multi-region with an ERP link. Build your timeline from the phases above and let the pilot results, not hope, set the cutover date.
Dormant records, unused custom fields, dead reports and automations nobody can explain. Classify each as archive or retire. Under GDPR’s storage-limitation principle, keeping personal data without a justified reason is also a risk, so ask privacy counsel about retention before you move old contact data.
Reconcile against criteria you set in advance: counts by object, open pipeline and forecast totals, associations, permissions, integration tests and user acceptance. Row counts alone don’t prove that revenue-critical relationships survived.
There’s no evidence in the migration guidance we’ve reviewed that it will. AI is more likely to change how data gets entered, scored and used, and it depends on clean, governed data. A well-planned migration prepares you for that rather than delaying it.
Subscribe to our Insights: Expert productivity tips in your inbox
You'll receive 1-3 emails per month. Your data stays private, always.
CRO & Co-Founder with Sales Label Consulting
Watch our Sales Mates Podcast
October 22, 2026 - 9 min read
Read article Read articleOctober 21, 2026 - 9 min read
Read article Read articleOctober 20, 2026 - 9 min read
Read article Read articleOctober 19, 2026 - 10 min read
Read article Read article